Collecting real-time fraud signals from a global creator network—how do you actually integrate that into your workflow?

I’ve been experimenting with something that sounds simple in theory but is messy in practice: pulling live signals from different parts of our creator ecosystem to feed into our fraud detection.

What I mean: we have brands reporting on creators they’ve worked with, we have other agencies flagging suspicious accounts, we have creators themselves sometimes calling out scams or fake partnerships. All of this is happening in real-time, but it’s scattered. Some of it’s in Slack channels, some is in emails, some is just word-of-mouth at conferences.

The thought is: what if we systematized this? Built a single place where signals come in, get structured, and then get fed into a risk assessment that actually influences which creators we vet further or which campaigns get an extra layer of review?

I started tracking this manually for a while—literally a Google Sheet where I’d log things like “follower spike of 15K in two days,” “engagement dropped 50% after month three,” “brand reported undelivered promised posts.” After a few months, I had about 200 entries. Some patterns emerged. Certain account behaviors really did correlate with problems later.

But manual tracking doesn’t scale. And you can’t rely on people remembering to log things. So the question is: how do you actually build a workflow where signals flow in automatically or with minimal friction, and then get processed in a way that’s actually useful?

Are you collecting signals like this? And if so, where’s the biggest bottleneck in your integration process—is it getting the data in, cleaning it, or actually acting on it?